CGenie - Professional CG Textures

For the last 25 years or so there has been an ongoing battle. On the one side you have the ambition and creativity of the CG artist and on the other you have the innovation and invention of the world of science and technology. You yourself have probably played your part in this, probably without consciously realising it - every time you render a scene you are looking to produce something better than the last render, and every time you push the limits of your hardware. It's an engagement driven entirely by the artists, who have raised the bar again and again. Advances in computing mean more detail, more realism can be achieved by the artist, but the artist simply devours that freed resource and demands even more.  As software and artistic ability has galloped ahead in recent years, the question is a whether this pace can continue or whether the direction of invention may be turning.

The March of Science

Although not the first to theorise the process (that would go to an IBM researcher named Appel in 1968 who published Some Techniques for Shading Machine Renderings of Solids), Turner Whitted is widely attributed as the father of raytraced rendering. In 1980, he released his now legendary paper An Improved Illumination Model for Shaded Display. He described an algorithm which allowed rendering of shadows, reflections and refractions. Yet what really made his name, was the simplicity. His method was so simple it could be summarised as three functions and around a dozen lines of pseudo code¹. Even today, the chances are your 3d rendering package is based upon these few simple lines.

This simplicity created a bridge. Whereas before raytracing had been the preserve of science and $400,000 supercomputers, it now appeared plausable that it could enter the professional market, and indeed the home.

Enter the Artist

In November of 1986, Commodore released a demo of a juggler juggling three mirror balls in a continual loop. It was full colour, 24 frames with even a matching 'clink' sound effect. The world was in awe. It sounds perhaps a little ridiculous now, but this was radical science then. You only need to look at the tone of articles written about it.  The May/June edition of AmigaWorld triumphantly heralded 'Behold the robot juggling spheres'. It was even used by Tom Petty in a music video!

Not only was rendering now possible to professionals, it was stuck to the covers of magazines sent to homes around the world. If there ever was a watershed moment, this was it. The fact that this was now an accessible application for the home user, meant that users could now start experimenting with the limits of the program. As the now-famous juggler shows in its rudimentary form, this was no longer about calculating pathways of light rays. It was about being creative.

The ambition of art

Now whilst all this was going on, good ol' Moore's Law was steadily continuing the explosion of processing power. At this point most users were looking at around 500k transistor's on their processing cores. So you'd think today with around 2000 times more power, we'd be raytracing in realtime and everything would be possible. But we're not.

I'd wager that we're waiting just as long for renders as Eric Graham was waiting for his juggler to appear, if not longer.

The reason is clear, the average render today is massively more detailed, more realistic and more creative in almost every fashion. Professional CG artists are no longer striving to achieve reality, they are toying with it. They are relied on by the film, games and music industry to create reality.

Yet there is a catch.

Unreal reality

Here's a little quiz from The Long Tail that highlights the point perfectly:

Q: On 1995 computer hardware, the average frame of Toy Story took two hours to render. A decade later on 2005 hardware, how long did it take the average frame of Cars to render?

A: 30 minutes
B: 1 hour
C: 2 hours
D: 15 hours

To spoil your fun guessing, I'll tell you. It's D. This is strange, though to most of us this is sort of expected given our experiences. The difference in realism between Toy Story and Cars isn't so great, it's certainly not 300 times more realistic. Yet with 300 times more processing power, we're actually taking more than seven times longer to render!

Per Christensen from Pixar explains the core differences in his powerpoint presentation - in short they were cheating. You will know yourself that most of 3d rendering is about finding workarounds and ways to balance render time with realism. With Toy Story they achieved it using scanline rendering, shadow maps, reflection maps. In Cars they used more of the real McCoy, the brute force-style raytracing - the no nonsense, CPU crunching approach that says 'take what happens to those trillions of light hitting each millimeter of our eye in the real world and just copy it'. But it's important to note that whilst the approach and method were opposed, the end result was not that radically different. The massive increase in processing power allowed them to make some steady gains in realism, but nothing transformational.

The future ahead

And that's where we are today. The 90s and early 2000s were defined by rapid and remarkable gains in realism - you may remember each of those releases of 3d Studio Max, Maya, the first sight of vray, vue d'esprit, Afterburn, where it really added a dimension that simply wasn't possible before. That seems to have slowed. It seems to me the next decade or so will simply have less scope to be about making things more impressive to look at. Go on, take a scene in Max or Maya and throw a thousand boxes into it. Now render with scanline and then render with raytrace. Not a massive difference really? Take a look at some of the great CG artists' work - what major improvement could they make?

Whether you're using Max or Maya, Modo or Cinema, Blender or Softimage. All are more than capable of producing stunningly realistic art.

We have already seen somewhat of an interface renaissance with the new intuitive interfaces introduced by the likes of Z-Brush, MudBox, Modo and countless others and the precedent has been set. The phrases you hear artists uttering are about it being 'intuitive', or 'easy to pick up' or 'flexible' or 'rock solid', not it's 'realistic' or 'lifelike' as people did of the juggler.

Artists want to be able to pick up and run with software immediately. They don't want to wade through manuals and technical support. Usability will be a key driver and, so it seems, realism is no longer a benchmark, it has become an expectation.